IEEE TRANSACTIONS ON EDUCATION 1 Integrating Mobile

IEEE TRANSACTIONS ON EDUCATION 1

Integrating Mobile Robotics and Visionwith Undergraduate Computer Science

Grzegorz Cielniak, Nicola Bellotto and Tom Duckett

Abstract—This paper describes the integration of robotics edu-cation into an undergraduate Computer Science curriculum. Theproposed approach delivers mobile robotics as well as coveringthe closely related field of Computer Vision, and is directlylinked to the research conducted at the authors’ institution.The paper describes the most relevant details of the modulecontent and assessment strategy, paying particular attention tothe practical sessions using Rovio mobile robots. The specificchoices are discussed that were made with regard to the mobileplatform, software libraries and lab environment. The paper alsopresents a detailed qualitative and quantitative analysis of studentresults, including the correlation between student engagementand performance, and discusses the outcomes of this experience.

Index Terms—Robotics education, robot vision, robot pro-gramming, Student as Producer.

I. INTRODUCTION

This paper describes the integration of robotics educationinto the undergraduate curriculum at the School of ComputerScience, University of Lincoln, UK. Rather than teachingrobotics in isolation, the subject is tightly integrated withother disciplines, in particular computer vision, as well asthe ongoing research activities in the School. The proposedapproach to robotics teaching features a holistic combinationof theory and practical work, including the programming ofvision-guided mobile robots in an open-ended assignmentloosely based on the popular RoboCup football tournament.This strategy also reflects the current policies of the institutionon research-informed teaching and the Student as Producer [1],a university-wide initiative supported by the UK Higher Edu-cation Academy.

Student as Producer “restates the meaning and purpose ofhigher education by reconnecting the core activities of univer-sities, i.e., research and teaching, in a way that consolidatesand substantiates the values of academic life” [1]. It does notdictate strict rules or policies but rather acts as a frameworkthat encourages thinking about educational processes withthe student engagement and active participation in mind. Thecore idea of the initiative is rooted in constructivism, whichalso inspired Papert’s constructionism [2]; it was Papert whoadvocated the use of technology in the learning process. Thisdirect link makes Student as Producer an especially attractiveframework in the current context, inspiring some of the choicesmade and the decision to involve students at various stagesof preparation of the Robotics module (e.g., the selection

G. Cielniak, N. Bellotto and T. Duckett are with the School of Com-puter Science, University of Lincoln, UK. e-mail: {gcielniak,nbellotto,tduckett}@lincoln.ac.uk

of software and hardware platform, creation of assessmentscenarios, etc.).

This paper presents the experience gained from conductingthese activities, the choices made during preparation of thetheoretical and practical work, the design of assignments, andvarious ways of assessing student performance.

II. RELATED WORK

In recent years, the use of robotics in education has gaineda significant interest among robotics experts and educators,illustrated by a number of funded projects (e.g., TereCop [3],Roberta [4]), workshops and conferences (e.g., [5], [6]). Dueto its interdisciplinary nature, robotics is often delivered incombination with other complementary subjects like electron-ics, mechanical engineering, computer vision [7], and the like.However, robotics is not only being taught as a specialist sub-ject to a narrow audience of future robotics specialists [8], butis being increasingly used as a tool for improving knowledgeof other subjects, developing transferable skills, enhancingstudent engagement in science, and more. This is closelyrelated to the recent increase in the availability of affordablerobotics platforms but is also due to the potential attractivenessof robotics technology to younger generations.

Previous work in educational robotics reports teaching activ-ities at different educational stages, including the primary, sec-ondary [9] and tertiary level [10]. These activities, dependingon the level, address various learning aspects including manualand communication skills [11], and scientific methods such asmeasurement, calculations, problem-solving, etc. Robotics isalso being used to improve understanding of other subjectslike maths, physics and engineering (e.g., [12]) and in generalfor improving student engagement in science subjects, [4].However, the efficient use of robotics tools requires specialisttraining for teachers themselves, especially those involved inthe early stages of education. While for more technically-oriented teachers this comes naturally, many others feel intim-idated by the complexity and the practical skills required [13].Educational robotics is also being used to address social issuesincluding gender balance, disabilities or disadvantaged socialbackgrounds (e.g., [11], [14]).

While there are many comprehensive sources covering theo-retical aspects of robotics (e.g., [15], [16], [17]), the main chal-lenge lies in the preparation of appropriate practical activities.The main difficulty comes from the fact that both pedagogical(e.g., specific age, or learning objectives) and technical aspectshave to be considered. A key decision related to the latter isthe choice of robotic platforms, taking into account reliability,


Lecture Topic C1 C2 C3

A1 Intro. to Computer Vision ◦ ◦A2 Intro. to Linear Algebra and MATLAB •A3 An Overview of Pattern Recognition • ◦A4 Spatial Processing and Filtering • ◦ ◦A5 Color Image Processing • ◦ ◦A6 Morphological Image Processing ◦ ◦ •A7 Image Segmentation I ◦ ◦ •A8 Image Segmentation II ◦ ◦ •A9 Image Representation and Description •A10 Pattern Classification •

B1 Introduction to Robotics ◦ ◦B2 Robot Programming in C# •B3 Actuators and Sensors • •B4 Robot Vision • ◦B5 Robot Control • •B6 Robot Behaviors • •B7 Control Architectures I ◦ •B8 Control Architectures II ◦ •B9 Navigation Strategies •B10 Robotic Map Building •

TABLE IMAIN TOPICS IN SEMESTER A (COMPUTER VISION) AND SEMESTER B(MOBILE ROBOTICS), WITH INDICATIVE COVERAGE BY ASSESSMENT

(• = HIGH, ◦ = LOW), DETAILED IN TABLE II.

Assessment Weighting Description Semester

C1 30% Image Processing in MATLAB AC2 30% Vision-based Robot Control BC3 40% Written Examination A + B

TABLE IITHE THREE ASSESSMENTS FOR THE MODULE INCLUDING WEIGHTINGS.

ease of use, maintenance, and so on. Thanks to the increasingnumber of affordable robots available on the market, thischoice is somewhat easier (see [18] for an extensive reviewof the available robotics platforms). In addition, affordablerobot kits have become popular toys and gadgets, furtherstimulating interest in robotics among younger generations.There are several popular scenarios that are employed forpractical activities including robotic football, home services,rescue missions [19], [20], or Braitenberg vehicles [21]. Thescenarios are acting in this case as micro-worlds [2] in whichthere are well understood objectives and requirements, as wellas potential for experimentation and discovery.

III. MODULE DESCRIPTION

Robotics at the Lincoln School of Computer Science isdelivered primarily in a study module called Computer Visionand Robotics. The module is targeted at the third-year (i.e.,final-year) undergraduate students due to its advanced content,its assumption of significant programming skills and its focuson mobile robots as complete systems, as well as on theircomponents. The module is compulsory for students on theComputer Science program, and optional for students studyingother programs at the School (Games Computing, ComputerInformation Systems, and Web Technology). There are severalrelevant modules that students undertake in earlier years ofstudy which provide the necessary background in program-ming and basic knowledge of Artificial Intelligence (AI),

Sto

rage

Cab

ine

ts

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

PC

NET

Exp

erim

en

tal A

rea

Com

pu

ters

+

Netw

ork

ing

Robots

Docking Stations

Football Pitch

Fig. 1. The layout of the computer lab used for practical sessions (left) anda snapshot taken during one of the sessions (right).

including Software Development in year 1, and AdvancedSoftware Development and AI in year 2. In addition, studentshave the option to study robotics as part of their self-guidedIndividual Study Project in year 3, in parallel with this taughtmodule.

Computer Vision and Robotics consists of two distinct parts:computer vision fundamentals are covered in the first semester(A) and robotics is covered in the second semester (B) – seeTable I for a detailed list of topics, and Table II for theirweightings. The computer vision part has a particular emphasison pattern recognition applications, hence the early overviewon this topic, but otherwise follows a fairly standard selectionof topics in digital image processing from a well-knowntextbook [22]. The weekly one-hour lectures, with additionalsupporting materials, are accompanied by a weekly two-hourworkshop in which the students learn to program in MATLABusing the Image Processing Toolbox (IPT) and the correspond-ing textbook [23]. Many techniques and concepts learned inthe computer vision part of the module are directly relevantand applicable to the robotics content. Indeed, robot vision is aprimary focus of the practical sessions and the assignment taskin the second semester. Each semester is individually assessedthrough a practical assignment (see Section IV) and there isa final written examination covering the theoretical content ofboth semesters. The robotics part of the module includes 12weeks of lecture delivery (comprising ten major topics plustwo weeks for supporting activities) and practical workshopsessions. The topics covered include the main challenges ofrobotics, robotic components, relevant software libraries, robotvision and control, robotic architectures, navigation strategiesand map building. The first set of lectures provides manypractical examples related to the robotic platform used, so thatthe students can experience a direct link between theory andpractice. The material covered also includes examples fromresearch conducted in the School to illustrate fundamentalproblems of perception (e.g., detecting people) and control(e.g., navigation), though the focus is mainly on “textbook”science where possible [15], [16], [17].

A. Practical Sessions

The robotics part of the module features 12 two-hour longworkshop sessions where the students have access to therobotic equipment. The sessions take place in a dedicatedcomputer lab with storage facilities and necessary space forexperimentation with the robots, Fig. 1. With this arrangement,all frequent tasks such as unpacking, setting-up, charging, and


Fig. 2. Rovio, the mobile robot (left) and the robot’s web interface (right).

so on can be carried out efficiently and smoothly withoutlimiting the amount of time for interaction and work withthe robots. In addition, the students can access the lab out-side the workshop hours, when it is not reserved for otheractivities. This arrangement appeared to be very popular inthe most recent offering of the course, especially towards theassignment hand-in date (!). The first four workshop sessionsare designed as introductory tutorials where the students learnabout the basic components of a robot, relevant softwarelibraries and programming principles. The tasks include stu-dents writing their own object detection algorithms, developingvisual feedback controllers, implementing simple behaviorsor using Finite State Machines for sequencing more complexbehaviors. The tutorial tasks were designed to provide a solidknowledge base as well as preparing necessary componentsneeded for developing the following assignment task (seeSection IV-B). The preparation of these teaching materialswas assisted by a Master’s degree student who had studiedthe module in the previous academic year, as a ten-weeksummer project supported by the university in connection withthe Student as Producer program [1]. The university’s policyof research-informed teaching is also aided by the part-timeemployment of current Ph.D. students in the area of visionand robotics as demonstrators during the workshop sessions.

B. Choice of Robot Platform

A number of choices had to be made regarding the platformand software libraries. Since the targeted students are exposedmostly to programming in C# in their previous years of study,it was decided to adopt this programming language for thepractical sessions. The students are provided with a simplesoftware wrapper directly implementing robot commands asspecified by the Rovio API document. An alternative wouldbe to rely on existing robotic suites such as Player/Stage,ROS or Microsoft Robotics Developer Studio; however, such achoice is perhaps more preferable for larger scale, long-termprojects as it requires substantial effort and time to becomefamiliar with these frameworks. At the same time, the studentslearn what to expect from a commercial platform in termsof software support and how to extend the functionality tomeet their own needs. The wrapper library [24] was developedfollowing object-oriented programming guidelines so that thestudents can apply the concepts learned in previous years toa real physical platform. The recommended image processinglibrary is AForge [25], which provides a well structured anddocumented functionality including a variety of image filtersand object detection algorithms. Ideally, OpenCV [26] would

Feature Rovio Roomba NXT

Affordability • • ?Maintenance/charging • ? ◦Set-up difficulty • ? ?Documentation ? ? ?Community Support ? ◦ •Software components/libraries • • •Quality of components ? ? •Vision support • ◦ ◦

TABLE IIIA COMPARISON OF ROBOTIC PLATFORMS USED FOR TEACHING ROBOTICS

IN THE CLASSROOM (SUPPORT: • = GOOD, ? = FAIR, ◦ = POOR).

be preferable due to its overall functionality, maturity andsupport. However, the existing C# wrappers are inefficient andrely on non-safe use of the language that could hinder thelearning process.

Rovio by WowWee [27] is an affordable mobile platformequipped with a set of sensors including a color camera(640x480 pixel resolution, 30 fps, RGB) mounted on a movinghead, odometry, infrared global navigation sensor and aninfrared obstacle detector (see Fig. 2). The omni-directionaldrive configuration with omni-directional wheels enables holo-nomic movement in all directions. The communication withthe robot is realized through wireless Ethernet; the on-boardcomputer (ARM-based architecture) runs a web-server thataccepts requests from and sends information to the remotePC, which can be programmed to realize different custombehaviors. There are several alternative robotic platforms thatare popular in delivering robotics courses, including Roombaand LEGO NXT. Table III presents a subjective comparisonof the popular robotic platforms based on the delivery team’sexperience of using them for teaching robotics in the previousyear.

While Roomba and LEGO NXT’s many attractive features(e.g. reconfigurability, direct control of wheels) have madethese platforms very popular at other institutions (e.g., [28],[29]), there are several reasons that make Rovio an idealplatform for the delivery of robotics in the proposed context.First, the robot is equipped with a color camera that isdirectly accessible from the software and can be used as theprimary source of sensory information in the workshop tasks.Other platforms (Roomba, NXT) require additional hardwarecomponents and solutions to enable video streaming. Theplatform is affordable – currently, it costs less than a singleRoomba robot or a standard Lego NXT set. This makes itpossible to purchase a large number of units that can be usedindependently by individual students in relatively large groups,which is very important for maintaining student engagementin the workshops. This also makes Rovio a more expendableplatform, as a broken or faulty unit can be easily replaced. Asimple and sturdy design, together with a recharging station,result in a straightforward and easily maintained set-up thatminimizes the time overhead for the preparatory tasks requiredbefore each session. The robot features a convenient com-munication interface through wireless Ethernet, enabling easysetup in existing computer laboratories (there are no driversrequired, and no additional dongles, connections or the like)


and a straightforward API that simplifies development of thestudents’ own robot behaviors, with minimal dependencies onother software libraries. Each Rovio robot in the lab is assignedits own IP address, and can be accessed directly through anordinary web browser (an important consideration when tryingto work with large groups of students).

The frequent and intensive use of the Rovio robots alsorevealed some limitations that might have not been immedi-ately obvious. Among them, the odometry sensor is unreliable,the global navigation sensor is unusable in a multi-robotenvironment, the camera image is of low quality, there can beconnection and bandwidth problems which are partially causedby the existing network facilities, and so on. On the other hand,these limitations helped the students to understand some ofthe fundamental issues and trade-offs in robotics and otherembedded systems, including networking, real-time control,sensor noise, interaction of multiple complex systems, etc.

IV. ASSESSMENT

The assessment strategy involves three separate compo-nents. Details of the topics covered by these three assessmentsare shown in Table I. The main objective of the first and secondcomponents, “Digital Image Processing in MATLAB” (C1)and “Vision-based Robot Control” (C2), is to assess studentpractical work during Semester A and B, respectively; themarks for these two components mostly reflect the functional-ity of the developed systems. The last component (C3) is thefinal (closed-book) written exam, designed to assess studentcomprehension of theoretical topics from the whole academicyear. Each component carries a weighting, Table II, which isused at the end of year to compute the overall module mark.

A. Digital Image Processing in MATLAB (C1)

The scope of this component is to assess knowledge andunderstanding of various aspects of computer vision, in par-ticular those related to image processing. This component issubdivided into three workshop tasks with increasing levels ofdifficulty, assessed at regular intervals throughout the semester,titled “Binary Images & Introduction to Pattern Recognition”,“Intensity Transforms & Spatial Filtering” and “Color ImageProcessing & Face Detection”. The deliverables for each taskconsist of a brief report describing the approach used to solvethe problem and the results obtained, accompanied by thecorresponding MATLAB source code. The latter is furtherdemonstrated in a follow-up workshop session, where studentsare required to answer specific questions about their ownMATLAB implementations.

B. Vision-Based Robot Control (C2)

The aim of this assignment is to evaluate competence in twomajor learning outcomes: the application of computer visiontechniques to solve practical problems, and the application ofAI control methods to mobile robotics. The assignment, whichbuilds upon computer vision expertise from the first semesterand some knowledge of AI from the previous year, is inspiredby the RoboCup competition [19].

The students are asked to design and develop a simplifiedversion of robotic football. The students can select one typeof player from a given list of striker, midfielder, defenderand goal keeper. The game features a uniformly colored balland a playing field consisting of a rectangular enclosure withdistinctively colored goals installed in the computer lab, Fig. 1.The minimum requirement for player functionality includesa ball searching behavior and a striking/defending behavior(which could be implemented as a reactive steering behaviorusing a proportional controller), depending on the player type.Students are free to choose the vision and robot controlalgorithms. Extra credit is given for developing additionalcomponents, including, but not limited to: 1) an enhancedobject detection system for learning object appearance orusing multiple cues; 2) behavior coordination for derivingsophisticated game strategies and implementing awareness ofother game objects, like goals and opponents; and 3) incorpo-ration of the above information into the player behavior. Thesubmission of this assignment includes a short technical reportdocumenting the design, implementation and evaluation of theproposed functionality as well as the developed source code. Inaddition, it is compulsory for the students to demonstrate theirwork during a separate workshop session after the submission,so the technical achievement and originality of the submittedwork can be assessed and graded.

C. Exam (C3)

The exam covers theoretical topics from both parts of themodule, and is particularly designed to evaluate students’critical assessment of the major topics covered. The exam istaken at a predetermined time and location, a few weeks afterthe end of Semester B, and must be completed within threehours. The students have to answer four questions, two fromthe three Computer Vision section questions, and two from thethree Robotics section questions. There are no programmingtasks in this final assessment. Instead, students are required todemonstrate their understanding of the systems and algorithmscovered in the two semesters (see Table I for an overview), andto compare various solutions to computer vision and roboticsproblems.

V. STUDENT PERFORMANCE

In the academic year 2010/11, the student cohort consistedof 18 Computer Science students for whom the module wascompulsory, four Games Computing students and one WebTechnology student who chose the module as an option, giving23 students in total. This is a relatively low sample andtherefore the reported results are anecdotal to some extent andmight not capture all characteristics of the course delivery.

A. Overall Results

Fig. 3 presents the comparative distribution of marks foreach assessment component, and sample correlation coeffi-cients between marks for each pair of assessment components.While the practical sessions proved to be popular and the stu-dents received relatively good marks for both assignments (C1


0 10 20 30 40 50 60 70 80 90 1000

1

2

3

4

5

6

7

marks

frequency

C1

C2

C3 C2

C30.62 C1

Fig. 3. Distribution of marks obtained for different assessment components(left) and mark correlation between each pair of assignment components(right).

C1 C2 C3

Lectures 0.82 0.63 0.52Workshops 0.70 0.55 0.37

TABLE IVSAMPLE CORRELATION COEFFICIENTS BETWEEN THE MARKS OBTAINEDFOR THE DIFFERENT COMPONENTS AND THE ATTENDANCE RECORDED

FOR THE LECTURES AND WORKSHOP SESSIONS.

and C2), the theoretical examination (C3) results were lowerthan expected – see the low correlation coefficients betweenthe exam and both practical assignments. These figures mightindicate a better engagement of the students in the practicalpart of the course, but perhaps also a wider problem ofstudent comprehension of theoretical material observed acrossall programs at the School.

To analyze the results further, attendance data was consid-ered as a basic indicator of student engagement, Table IV.The results indicate that attendance at lectures and workshopsshowed a strong correlation with performance on the Com-puter Vision assignment (C1), although less so in the secondsemester. Perhaps surprisingly, the attendance data showed thelowest correlation with the marks obtained in the exam.

Further analysis included the correlation between marksobtained in Computer Vision and Robotics and other rele-vant modules taken by students during their course of study.Table V presents sample correlation coefficients in the formof a matrix between different modules including a first-yearmodule, Software Development (SD); second-year modules,Advanced Software Development (ASD) and Artificial Intel-ligence (AI); and third-year modules Computer Vision andRobotics (CVR), Advanced Software Engineering (ASE) andIndividual Study Project (ISP). While it can be seen thatthe correlation between CVR and other third-year modules isstronger than with modules from earlier years, it is interestingto note that CVR is ranked as the module most correlated withASE, ISP and ASD, i.e., the modules where programmingskills play a prominent role.

B. Robotics Assignment – Observations

The robotics assignment had clearly defined minimum re-quirements but fairly open goals, which encouraged exper-imentation and exploration. This resulted in a number ofexceptionally good submissions which included many featuresbeyond the standard specifications. All the best submissions

CVR ASE ISP ASD AI SD

SD 0.34 0.41 0.38 0.37 0.53 1.00AI 0.49 0.40 0.49 0.62 1.00

ASD 0.62 0.54 0.54 1.00ISP 0.87 0.87 1.00ASE 0.94 1.00CVR 1.00

TABLE VSAMPLE CORRELATION COEFFICIENTS BETWEEN COMPUTER VISION AND

ROBOTICS AND OTHER RELEVANT MODULES (SEE SECTION V-A FORFULL LABEL DETAILS).

(marks greater than 70%) had a clearly defined individualfocus and explored different issues and directions. Someexamples of the outstanding achievements presented by thestudents included:

• object detection: a multi-stage image processing pipelineincluding cascaded segmentation in different colorspaces, morphological operators for noise filtering, theuse of edges as additional features, and histogram-basedtuning and learning of image filter parameters;

• control architectures: a hybrid architecture combiningreactive and deliberative approaches, complex behaviorsequencing models (e.g., hierarchical FSM), a predictiveball search behavior, multi-threading and synchronizationfeaturing customized threading queue mechanisms andoff-line system development using prerecorded data sets;

• system evaluation: rigorous quantitative evaluation in-cluding switching time analysis of behavioral models anddedicated testing scenarios, consideration of trade-offs(e.g., speed vs. accuracy) for robot controllers.

The above list of topics indicates that the students wereable to apply not only knowledge learned in the previoussemester (for example, object detection) but also that acquiredin other modules including AI, Software Development andSoftware Engineering. Many of the listed techniques requiredthe students to research sources other than the recommendedreading material. On average, the students with the highestgrade referenced two items from the recommended readinglist, three items citing other books, journal publications andconference papers discovered through individual research, andone item citing software or hardware.

VI. DISCUSSION AND CONCLUSIONS

It has been pointed out previously that the goal of “educa-tional robotics in general, is not precisely to teach learners tobe robotics experts but to develop the essential competences tobe successful in the present world” [28]. The authors believethat the module presented here gives students vital expertise inareas that are otherwise not strongly covered in the “standard”computer science topics, such as dealing with complex sys-tems at a systems (and systems of systems) level, combininghardware with sensing and control software, understanding thepracticalities of real-time systems, understanding the inherentuncertainty in the real-world as perceived through sensors(sensor noise), and applying programming methodologies inpractice. The students demonstrated a very high level of


engagement, spending a significant amount of time solving therobot football task required for the assignment. This resultedin a number of exceptional submissions with very originalfunctionality that went beyond the assignment brief (incor-porating such features as threading queues, speech synthesis,etc.). While many students struggled with the evaluation partof the assignment (most likely due to their time manage-ment), many of the technical issues and platform shortcomingswere identified by the students (connection and bandwidthproblems, granularity of the movement commands, limitedodometry, changing light conditions, etc.). The laboratoryspace encouraged cooperation, support and competition indeveloping individual solutions. The module proved to bepopular among the students and there was an increased intake(44 students) in the academic year 2011/12. Some pedagogicalissues to be addressed in future extensions of this workinclude, for example, time requirements for solving practicalassignments and scenarios alternative to RoboCup. Furtherdevelopment plans for the module include: a common softwarerepository to teach code maintenance and development inteams, extensions to the software environment, multi-robotscenarios, and a greater involvement in the Student as Producerinitiative [1], including recruiting more student helpers andbuilding stronger links with the student Computing Society.A comparative study with the previous year and with otherinstitutions delivering similar content is also envisaged.

REFERENCES

[1] M. Neary and J. Winn, “The student as producer: reinventing the studentexperience in higher education,” in The future of higher education:policy, pedagogy and the student experience. Continuum, 2009.

[2] S. Papert, “What is Logo? And who needs it?” in Logo philosophy andimplementation. Logo Computer Systems Inc., 1999.

[3] D. Alimisis, M. Moro, J. Arlegui, A. Pina, S. Frangou, and K. Papaniko-laou, “Robotics & Constructivism in Education: the TERECoP project,”in Proc. of the 11th European Logo Conf., Bratislava, Slovakia, 2007.

[4] Roberta Project. [Online]. Available: http://www.roberta-home.de/en[5] International workshop: Teaching robotics teaching with robotics.

[Online]. Available: http://www.terecop.eu/TRTWR2012.htm[6] International robotics in education conference. [Online]. Available:

http://www.rie2012.eu/[7] G. Bebis, D. Egbert, and M. Shah, “Review of computer vision educa-

tion,” IEEE Transactions on Education, vol. 46, pp. 2–21, 2003.[8] B. A. Maxwell and L. A. Meeden, “Integrating robotics research with

undergraduate education,” in IEEE Intelligent Systems Magazine, 2000.[9] M. Kabatova, L. Jaskova, P. Lecky, and V. Lassakova, “Robotic activities

for visually impaired secondary school children,” in Proc. Int. WorkshopTeaching Robotics Teaching with Robotics: Integrating Robotics inSchool Curriculum, Riva del Garda, Italy, April 2012.

[10] J. Carter, S. Matthews, and S. Coupland, “Teaching robotics at thepostgraduate level: Assessment and feedback for on site and distancelearning,” in Proc. Int. Conf. on Robotics in Education, 2011.

[11] T. Urschitz and T. Carl, “Mission on Mars. Interactive robotics session,on video conference, through the use of Interactive White Board (IWB),”in Proc. Int. Workshop Teaching Robotics Teaching with Robotics:Integrating Robotics in School Curriculum, Riva del Garda, Italy, 2012.

[12] L. Riano and M. McGinnity, “Design and validation of a robotic systemto interactively teach geometry,” in Proc. Int. Conf. on Robotics inEducation. FEI STU, Slovakia, 2010, pp. 5–9.

[13] D. Alimisis, “Integrating robotics in science and technology teachertraining curriculum,” in Proc. Int. Workshop Teaching Robotics Teachingwith Robotics: Integrating Robotics in School Curriculum, Riva delGarda, Italy, April 2012.

[14] D. Catlin and S. Robertson, “Using educational robots to enhance theperformance of minority students,” in Proc. Int. Workshop TeachingRobotics Teaching with Robotics: Integrating Robotics in School Cur-riculum, Riva del Garda, Italy, April 2012.

[15] G. Bekey, Autonomous Robots: From Biological Inspiration To Im-plementation And Control, ser. Intelligent Robotics and AutonomousAgents. Mit Press, 2005.

[16] R. Siegwart and I. R. Nourbakhsh, Introduction to Autonomous MobileRobots. Bradford Company, 2004.

[17] R. R. Murphy, Introduction to AI Robotics, 1st ed. MIT Press, 2000.[18] M. Ruzzenente, M. Koo, K. Nielsen, L. Grespan, and P. Fiorini, “A

review of robotics kits for tertiary education,” in Proc. Int. WorkshopTeaching Robotics Teaching with Robotics: Integrating Robotics inSchool Curriculum, Riva del Garda, Italy, April 2012.

[19] M. Asada, R. D’Andrea, A. Birk, H. Kitano, and M. Veloso, “Roboticsin edutainment,” in Proc. IEEE Int. Conf. on Robotics and Automation,2000, pp. 795–800.

[20] A. D. Tonkonogui, P. Bell, A. R. Linse, and A. V. Mamishev, “IntegratingFIRST robotics program with university curriculum,” in Proc. Int. Conf.on Robotics and Applications. Anaheim, CA, USA: ACTA Press, 2007.

[21] E. Datteri, L. Zecca, F. Laudisa, and M. Castiglioni, “Explaining roboticbehaviors: a case study on science education,” in Proc. Int. WorkshopTeaching Robotics Teaching with Robotics: Integrating Robotics inSchool Curriculum, Riva del Garda, Italy, April 2012.

[22] R. C. Gonzalez and R. E. Woods, Digital Image Processing (3rdEdition). Prentice Hall, 2008.

[23] R. C. Gonzalez, R. E. Woods, and S. L. Eddins, Digital Image Process-ing Using MATLAB (2nd Edition). Gatesmark Publishing, 2009.

[24] RobotLib project. [Online]. Available: http://robotlib.codeplex.com[25] AForge.NET. [Online]. Available: http://code.google.com/p/aforge[26] G. Bradski, A. Kaehler, and V. Pisarevsky, “Learning-based computer

vision with OpenCV,” Intel Technology Journal, vol. 9, 2005.[27] WowWee Rovio. [Online]. Available: http://www.wowwee.com/

en/products/tech/telepresence/rovio[28] K. Pitti, B. Curto, J. Garcia, and V. Moreno, “NXT workshops: Con-

structionist learning experiences in rural areas,” in Proc. Int. Work-shop ”Teaching robotics, teaching with robotics”, Darmstadt, Germany,November 2010.

[29] E. Menegatti and M. Moro, “Educational robotics from high-school tomaster of science,” in Proc. Int. Workshop ”Teaching robotics, teachingwith robotics”, Darmstadt, Germany, November 2010.

Grzegorz Cielniak is a Senior Lecturer at the School of Computer Science,University of Lincoln, UK. He obtained his Ph.D. in Computer Science fromOrebro University, Sweden in 2007 and his M.Sc. in Robotics from WrocławUniversity of Technology, Poland in 2000. His research interests includemobile robotics, machine perception, estimation and tracking techniques,monitoring of humans and other species. Grzegorz is currently involved inteaching a number of undergraduate computer science subjects at Lincoln,including the module Computer Vision and Robotics.

Nicola Bellotto is a Senior Lecturer in Computer Science at the Universityof Lincoln, where he teaches a number of undergraduate and postgraduatemodules, including Computer Vision and Robotics and Operating Systems.His research interests range from mobile robotics to cognitive perception,including sensor fusion, Bayesian estimation and embedded AI. He holds aPh.D. in Computer Science from the University of Essex and a Laurea inElectronic Engineering from the University of Padua. Before joining Lincoln,he was a post-doctoral research assistant at the University of Oxford. Nicolahas also several years of professional experience working in the industry assoftware developer and embedded system programmer.

Tom Duckett is a Reader in Computer Science at the University of Lincoln,UK, where he is also Director of the Center for Vision and Robotics Research.He was formerly a docent (Associate Professor) at Orebro University, Sweden,where he was leader of the Learning Systems Laboratory within the Centerfor Applied Autonomous Sensor Systems. He obtained his Ph.D. in ComputerScience from Manchester University in 2001, M.Sc. with distinction inKnowledge-Based Systems from Heriot-Watt University in 1995 and B.Sc.(Hons.) in Computer and Management Science from Warwick University in1991, and also studied at Karlsruhe and Bremen Universities. He currentlyteaches artificial intelligence, computer vision and mobile robotics as part ofthe B.Sc. program in Computer Science at Lincoln.

Documents

IEEE TRANSACTIONS ON EDUCATION 1 Integrating Mobile